Partially supported by a NULab Seedling Grant.
Advanced capabilities in the collection and sharing of large-scale datasets promise to improve research and innovation, and so benefit societies at large. But this increase in the amount and reach of data is not unmitigated good. For example, promising results about the efficacy of a medical treatment can quickly travel far and wide in densely connected scientific communities. But this increase in the speed of discovery can come at a cost, when, as a consequence of focusing on these promising results, the scientific community ceases its exploration for, and so misses out on, potentially better alternatives. More generally, as existing research has shown, the impacts of increased data sharing depend on a variety of factors, ranging from structural properties of communicative networks to interpersonal relations between different agents (e.g., individuals or groups), to the norms and standards followed by these agents. What is more, we often face tensions between different values that are of both scientific and societal interest, such as the trade-off between the speed and reliability of discoveries. When designing and evaluating policies that seek to improve research and innovation practice, we need to closely examine their impacts through the lens of varied values, which might differentially matter to different stakeholders.
This project focuses on a value that, while critical to responsible research and innovation, has largely remained outside the purview of current works: the equitable distribution of the benefits and burdens of scientific and technological enterprises across different demographics. Closely attending to these considerations of justice and fairness is particularly important because many outcomes that appear beneficial at the level of the general population turn out to be deeply unjust upon the examination of their disaggregated impact on different sub-populations. To this end, this project develops a novel simulation model and a set of fairness metrics for disaggregating, examining, and evaluating group-specific impacts of different aspects of research and scientific communities. We put these tools to work to examine ethical and policy-related questions such as: How might governance proposals for enhancing the reliability of scientific research in the age of big data impact the equity of scientific enterprises? What distinct obligations might research and innovation communities face when they share heterogeneous data?
Sina Fazelpour, Faculty, Philosophy and Computer Science